Performance Considerations

Optimizing DSP performance in bbx_audio.

Key Metrics

Optimization Strategies

Pre-computation

Calculate once, use many times:

#![allow(unused)]
fn main() {
// In prepare()
self.coefficient = calculate_filter_coeff(self.cutoff, context.sample_rate);

// In process() - just use it
output = input * self.coefficient + self.state;
}

Cache Efficiency

Keep hot data together:

#![allow(unused)]
fn main() {
// Good: Contiguous buffer storage
audio_buffers: Vec<AudioBuffer<S>>

// Good: Sequential processing
for block_id in &self.execution_order {
    self.process_block(*block_id);
}
}

Branch Avoidance

Prefer branchless code:

#![allow(unused)]
fn main() {
// Avoid
if condition {
    result = a;
} else {
    result = b;
}

// Prefer
let mask = condition as f32;  // 0.0 or 1.0
result = a * mask + b * (1.0 - mask);
}

SIMD Potential

Design for SIMD:

• Process 4+ samples at once
• Align buffers to 16/32 bytes
• Avoid data-dependent branches

Profiling

Measure before optimizing:

#![allow(unused)]
fn main() {
#[cfg(feature = "profiling")]
let _span = tracing::span!(tracing::Level::TRACE, "process_block");
}

Common Bottlenecks

1. Memory allocation in audio thread
2. Cache misses from scattered data
3. Branch misprediction in complex logic
4. Function call overhead for tiny operations
5. Denormal processing in filter feedback